为提高盲提取算法在恢复短样本信号时的质量,本文提出基于时间结构的盲提取方法.该方法以多个延时量上统计得出的自相关平方和取代负熵来作为目标函数,并引入了输出信号与参考信号的近似测度和输出信号方差作为非线性规划的约束条件,结合拟牛顿迭代法,推导出新的单路盲信号提取迭代表达式.另外,本文还提出了衡量短样本程度的量化标准.仿真实验表明,不论所观测的各路信号长度如何,只要其包含短潜伏期的源成分,本文方法均能高质量地完成有用信号的提取.

To enhance the quality of the restored signal extracted from short-sampled paths,this paper proposed a novel blind extraction method based on time structure.By replacing negentropy with the sum of autocorrelation statistically obtained at multiple delays as the objective function,and introducing the closeness between output signal and reference signal as well as output's variance as the constraints of non-linear programming,besides incorporating the Newton-like iterative method,a novel iterative expression was successfully deduced to extract one-path signal.In addition,a novel criterion to evaluate the short samples is presented.Simulation showed that,whatever the length of observed multiple signals is,as long as the short-duration component is contained,the proposed method can extract the useful signal in high quality.